In the field of adaptive welding, many attempts have been made to automate the welding process, as evidenced by the tremendous body of art principally directed to locating and tracking weld grooves. The art ranges from simple tactile sensors, which provide a good deal of information on the location of the groove, but little about the groove's dimensions or shape, to the more advanced vision type systems, mimicking the human operators method of guiding the welding torch along the groove. In fact, the human operator appears to be the model, or at least the standard to which all machine welds are compared. Adaptive welders have shown an ability to weld many times faster, for longer periods of time, and with a consistency considered unachievable by their human counterparts. Given that the adaptive welders perform the functions of the human welder with a higher degree of accuracy and consistency, one may reasonably assume that adaptive machine welding is consistently of a higher quality than those welds performed by human operators. However, no connection has been suggested between the quality of a weld and the guidance point of the welding head within the weld groove. For example, guiding to the center of a symmetrical weld groove may provide a weld of superior quality, but implementing the same guidance scheme on a weld groove having a nonsymmetrical cross sectional configuration could result in a weld having less than ideal physical characteristics. Similarly, guiding the welding head at a constant distance from one of the groove edges may prove less than satisfactory in areas of groove width variation, where in the worst case, welding could actually be performed outside the groove. At best, welding will occur at the optimum position within the groove purely by chance and only momentarily.
The present invention is directed to overcoming one or more of the problems as set forth above by positioning the welding head at the optimum welding position and adaptively maintaining this optimum positioning during the welding process.